Adaptive Closed Loop OFDM-Based Resource Allocation Method using Machine Learning and Genetic Algorithm

TL;DR

This paper introduces an adaptive, machine learning-driven resource allocation framework for OFDMA-based LTE networks, utilizing genetic algorithms to optimize scheduler performance based on user demands and network conditions.

Contribution

It presents a novel ML-based adaptive scheduling framework that dynamically adjusts objectives, overcoming traditional trade-offs, and demonstrates its effectiveness with a GA-based scheduler.

Findings

Optimizes resource allocation based on user traffic patterns.

Enhances scheduler efficiency with ML and genetic algorithms.

Provides a flexible, generic framework adaptable to various scheduling strategies.

Abstract

In this paper, the concept of Machine Learning (ML) is introduced to the Orthogonal Frequency Division Multiple Access-based (OFDMA-based) scheduler. Similar to the impact of the Channel Quality Indicator (CQI) on the scheduler in the Long Term Evolution (LTE), ML is utilized to provide the scheduler with pertinent information about the User Equipment (UE) traffic patterns, demands, Quality of Service (QoS) requirements, instantaneous user throughput and other network conditions. An adaptive ML-based framework is proposed in order to optimize the LTE scheduler operation. The proposed technique targets multiple objective scheduling strategies. The weights of the different objectives are adjusted to optimize the resources allocation per transmission based on the UEs demand pattern. In addition, it overcomes the trade-off problem of the traditional scheduling methods. The technique can be used as a generic framework with any scheduling strategy. In this paper, Genetic Algorithm-based (GA-based) multi- objective scheduler is considered to illustrate the efficiency of the proposed adaptive scheduling solution. Results show that using the combination of clustering and classification algorithms along with the GA optimizes the GA scheduler functionality and makes use of the ML process to form a closed loop scheduling mechanism.